haldujai

> it becomes a contest of who has the longer runway

Which is what I said - frontier labs do not have a sufficiently larger war chest.

Respectfully, comparing the balance sheets of a frontier lab to Blackrock, Apollo, Softbank or Nvidia is laughable.

The neocloud will also easily rent out that unused compute and keep funding their war chest via Nvidia’s backstop. If Nvidia can’t pay then it’s all over and AI companies have been swallowed by credit markets.

Your hypothesis also ignores the fact that if frontier labs are not paying for compute it is because they have bad revenue, little cash available and no credit (no one is loaning money to OAI or Anthropic if they can’t pay their bills. They can barely get debt today).

> Regardless, the terms come down to positioning and bargaining power. Runway vs obligations. This seems to dramatically favor the labs.

No, the only way out is for OpenAI or Anthropic to claim in court that they would be insolvent if they are required to pay the bill in full. Otherwise this is a straight forward deal.

You are greatly overestimating their capabilities and underestimating how many substantially more well-funded and highly liquid players will be lining up to take their pound of flesh if OAI or Anthropic default on a bill.

I’d wager my life savings one of those two will go through liquidation.

> Do you have a war chest to drag out a legal battle and the ability to bankrupt your landlord

Their war chest is not sufficiently larger than the hyperscalers, Broadcom or PE financiers like BlackRock.

> The rules are different at this scale, and "you can't enforce a contract if the attempt will bankrupt you" is only one of the possible games.

Rather, the inverse of what you said and far less Machiavellian.

It is legally well established that take-or-pay contracts are unenforceable if it would bankrupt the buyer which is the more likely resolution in this case. However, this does not mean frontier labs just get to walk away. They would settle in court for a penalty that is less than the full contract price.

See: AMD/global foundries and pretty much every oil and gas or commodity deal that goes bust when market pricing makes extraction unfeasible.

Frugal when it comes to AI model development.

> China will become the world's reserve currency by 2032.

China doesn’t want this for the same reason that some Americans currently in power don’t want the USD to be the world’s reserve currency - it comes with an export and trade cost.

There are also many structural and political reasons why the Yuan can’t play this role not least of which are transparency, the peg and monetary supply independence.

While not the only ones, the NYT notoriously pushed a lot of false and poorly verified information from questionable sources as fact in an effort to be the first to publish. They subsequently issued a mea culpa.

https://www.nytimes.com/2004/05/26/world/from-the-editors-th...

No, just that “What if I just cheat to get the treat” is also something my lovable Boxer does when I hide rewards as well.

I suppose her energy is limited but even if infinite, and with caution, I estimate the chances of a mass extinction event due to Pickle taking over the world at < 1%.

Not explicitly, but hacking HF is within the scope of “solve this problem at all costs” + no/poor guardrails + infinite budget + unsolvable problem.

> The US produces about 70% more electricity per capita.

And consumers use 4x as much per capita. Industrial generation per capita China comes out ~2x

> industrial electricity prices in China are roughly 34% higher than in the US

For which industrial customer and where? Chinese compute hubs are on par to slightly cheaper on pure electricity costs.

Conversely the US makes it more expensive with interconnect and upgrade fees as well as hefty take or pay contracts.

A 1GW datacenter in VA for example would add 5-10c kWh and a 12 year take or pay deal

1. I don’t think that’s a very strong argument. OpenAI and Anthropic don’t buy the vast majority of GPUs they use they rent capacity.

Nvidia could just the same rent those GPUs out for inference and actually have way better margins than they do right now. Antitrust and putting all your eggs in one basket are why they don’t, similar to TSMC.

2. Neither do AI labs. See Anthropic buying TPUs, deploying with AMD. OpenAI on Maia, Cerebras, their own wafers.

About the same, 5-10, when you consider major (aka frontier) airlines.

Actually not a bad comparison. Both burn massive amounts of up front capital to protect an oligopoly in the hopes their commodity product eventually pays off.

> 1. The whole Burry argument is that AI hardware becomes obsolete faster. If aux hardware can be reused, that works against the argument.

Burry’s main argument is depreciation is being understated and the capex vintages will not be paid off before they are essentially useless. This can happen whether or not aux is reused.

> Total consumption drives more demand for the already supply constrained hardware.

Demand is the wrong metric.

Only number that matters is whether AI attributable revenue will be sufficient to pay back enough of each successive capex vintage (e.g. 750B this year, 1T next year, 1.2T in 2028) so that hyperscalers and neoclouds can either self-fund or continue to issue debt as bond markets are already straining and tax-payer backed sovereign debt is providing a high baseline. Otherwise they downgrade capex projections and the bubble pops.

Expensive compute needs expensive inference to justify 30-40B/year/GW of compute. There are many reasons why frontier API pricing which is what the industry is based on may not persist. It is also almost certainly the case that 2026 is the worst year of supply and demand mismatch to allow for 80%+ margins. HBF next year has the potential to single handedly pop the DRAM spot bubble.

> Can AI hardware market go bust? Sure it can.

This is the bear thesis. It is not that AI will crash or be useless.

> But being early is the same as being wrong in the investment market. When do you predict the bust to be?

Q4 27-Q2 28 is when the bill becomes due at the latest. There are sufficient financial levers left to buy time without returns until then.

> Google, Amazon, Microsoft, Meta are all buying as many Nvidia GPUs as they possibly can.

All of these companies have rock solid revenue streams and can easily swallow 500B of capex devaluation over time. Their buying of Nvidia today is not necessarily the indicator you are implying as there are strong competitive reasons to make the game more expensive for everyone else.

1. Not really, current valuations are priced for persistent 80%+ margins based on spot. If auxiliary hardware lasts longer (I.e. next gen GPU reusing the same shell) then that reduces supply pressure and spot prices.

2. Jevon’s paradox is about total consumption, not margins. Valuations are about margins (and their projections). Many coal mine owners went bust despite increased total coal consumption.

3. Source? Gemini for example is 70% on TPU. I have yet to see data on Maia-300 beyond Microsoft PR. Remember it doesn’t have to be better it has to be more cost efficient. The overwhelming majority of inference spend does not care if token output is 20% slower if it is 50% cheaper.

> Neoclouds may very well be Nvidia’s biggest customers and this probably what Nvidia wants.

What Nvidia needs. Whether neoclouds can stay competitive vs hyperscalers paying Nvidia tax is far from clear, particularly when inference margins compress.

1. The same isn’t necessarily true of the rest of the hardware stack which may be reused between accelerator generations.

2. You’re missing the “New Nvidia chips 10x B200, compute requirement grows less than 10*software improvements YoY -> buy less Nvidia.” Valuations are based on forward projections (>1T annual for NVDA) which can be revised down leading to a drop in valuation.

> If Amazon doesn't buy but Microsoft does

The big 3 all have their own proprietary accelerators. Meta is buying TPUs as well for now.

I would bet Nvidia’s major customers in 2 years are neoclouds and it seems that Jensen is making the same bet.

Paying customer at what price? My understanding of Meta’s request is API pricing per token (of course discounted for volume) for content moderation etc. which is a lower margin product than Gemini enterprise seats. They also prioritize internal training runs.

This doesn’t mean more compute at any price is worthwhile. It also doesn’t mean that the SpaceX compute deal would even be offered to Meta.

> I don't get this though: The theory is all these hyperscalers are simultaneously spending buttloads of money on CapEx to the extent it affects their stock price, and then they would lie about the demand to protect their share price.

It’s not lying - it’s optimistic revenue projections. Your Sam Altman point is an example, these RPOs are real but what’s questionable is whether the AI labs can generate enough premium token API revenue to actually pay those commitments. Today’s OpenAI annualized revenue estimate is only 40B. Will they actually be able to 10x that to pay those RPOs? I’m skeptical especially with offloading inference to cheaper models.

> Why would they do all that when they could just do nothing and keep their firehoses of existing business revenue untouched and maintain their stock prices on the upward trajectory they already were -- like Apple?

The hyperscalers with proven revenue streams and strong financials (Amazon, MSFT, Google, arguably Meta) benefit from making the game more expensive than everyone, will get at least 50% of their capex back from this peak supply/demand mismatch and maybe other than AWS could easily use any excess compute for internal needs.

It is very suspicious when they’re paying for GB300 more than they turn around and charge those out as a4x instances, it was announced a week before IPO and they have a 90 day exit option.

> The RPO can be anywhere from 3 - 6 years, sure, but even on an annual basis that’s like a hundred billion now.

OpenAI and Anthropic deals are mostly 5 year and Anthropic’s starts in 2027, accordingly these RPOs are sized for projected compute needs and run rate in 2027 not today.

The only way either lab could pay 1 year of RPOs today (~80B for anthropic and ~150B for OpenAI) is with a lot more debt or circular financing, the former of which is difficult in this market.

Everyone spending crazy money on capex right now says they have a crushing backlog and need more compute to protect their share price. I highly doubt the demand exists today at current prices if the big 3 hyperscalers magically had an extra 2-3GW of compute. It’s not like Anthropic and OpenAI are turning away customers offering to pay API pricing..

Backlog meaning RPO over 5 years. It’s not as if they would be able to collect 250B today from OpenAI and Anthropic if they were to have that compute.

In any case, my point is that the SpaceX deal specifically likely has ulterior motives.

A deal with a company Google has a share of, announced a week before their IPO, with very non-committal terms and ramp period protections delivered in one large block on short term notice priced likely at the high end of what Google charges for A4X instances anyway.

I don’t think this reflects desperation as much as strategy.

It hasn’t bent already? 2022-2024 certainly seemed far more exponential than 2024 to present.

No, that’s not my argument. Rather it is that a counterexample to a mathematical proof whether produced by human or AI may be a scientific discovery but it is not inherently new and brilliant by definition.

The case for AI is also weakened when the model is steered by an expert.

The comment more likely means the harness source was read, not memory from a previous run and the first few turns of bp35 appear to be a cold start.

Sure none of this is certain without the source.

I do believe the authors that this schema significantly improves over the base, particularly given that it took 22x simulated turns over 14 hours, which is moving the trial and error to context rather than to game. I also don’t doubt there is some contamination.

Regardless, the approach is sound and I do believe it would significantly improve scores, even if that was +20-30 over baseline (49% in this case) it does imply the benchmark is measuring the harness more than the model.

If you are referring to the Jacobian conjecture Claude only provided a counterexample, not a disproof, neither of which is necessarily a “new brilliant scientific idea”.

No but the session traces are available. It passes the sniff test considering how AGI-3 is scored and how this wrapper works.

For example on bp35 it took fable 290M and >12k simulated turns for 566 real turns and finish more efficiently than a human.

Regardless of the true score I think the takeaway is the benchmark measures the wrapper rather than the model.

https://huggingface.co/schema-harness

For frontier models, not local.

https://schema-harness.github.io/

> Say, Claude 7 creates a new, brilliant scientific idea every 1 out of 1000 times.

I would not hand wave that away - but Claude 5 feels closer to Claude 1 than the hypothetical Claude 7 you propose. I do not believe one can extrapolate LLMs that far ahead despite the very substantial progress so far.

Agreed. Berkeley / CXRO is the American beyond-NA research center not Albany NanoTech which doesn’t have a charter AFAIK. They are just beginning to install the EXE:5200 in Albany now..

They do have the xLight program which doesn’t have a true European equivalent but still is a strategic partner of ASML who have experimented with FEL in the past.

Claude, the model in the post.

Radiologist who does read shoulder MRI would like to add that over half the annotations are wrong, glaring mistakes in anatomy and cardinal direction which begs the question of how is it making these findings without knowing what it’s looking at (here’s a hint, it’s hallucinated based on reports it sees).

Ultrasound is very operator dependent. Shoulder ultrasound is very hard. Visualizing the labrum let alone detecting pathology is very very hard and you will miss huge chunks of it due to limited windows.

Ignoring all of this, there are few sub specialist radiologists in the world who could theoretically do this and if you were to pay for their time it would cost more than a highly reproducible and easy to get MRI.

Half baked images? No one said that is impossible.

I am skeptical of any brain ultrasound claim that doesn’t use skull correction which requires a CT scan.

Very large chunks of vasculature and major arteries are missing in the images they provided. Just because it’s pretty and colorful it doesn’t mean it’s useful.

Perhaps it will one day, but this doesn’t prove much so far. There are several physical challenges to using ultrasound.

Depends for what reason. I trained in Canada. Studies are triaged priority 1 through 4 in most provinces. Nowhere in Canada is a high priority MRI waiting 2 months.